PT104902A - METHOD OF NON-DESTRUCTIVE AND NON EVASIVE INSPECTION OF VEGETABLE MATERIALS BASED ON THE USE OF ELECTROMAGNETIC RADIATION - Google Patents

Abstract

This invention describes a method for the monitoring of intrinsic parameters of natural cork reels based on a non-invasive and non-destructive method, for example the RX spectroscopy, with a view to segregating and categorizing objects according to the ratio of these parameters WITH RELEVANT CHARACTERISTICS FOR THE PERFORMANCE OF THE SEALANT, AS FOR EXAMPLE THE PERMEABILITY TO GASES. THIS INVENTION DESCRIBES STILL INFORMATION PROCESSING ON CORK PHYSICAL-CHEMICAL PARAMETERS DIRECTLY FROM THE DATA COLLECTED ON THE PRODUCTION LINE. IT IS ESSENTIALLY CHARACTERIZED BY UNDERSTANDING: MEANS FOR THE ACQUISITION OF SIGNAL; AN AUTOMATIC ALGORITHM OF DETECTING THE OBJECTS IN THE RADIOGRAPHIC IMAGE; AN ALGORITHM FOR THE DETECTION OF THE GROWTH TABLE ON WHICH SIGNAL TREATMENT IS MADE - THE SIGNAL PROCESSING IN REAL TIME; SOFTWARE FOR THE RECOGNITION OF PATTERNS OF BEHAVIOR RELATING TO THE VARIABLE TO QUANTIFY; CATEGORIZATION BY THE SIGNAL RECEIVED BY SENSORS IN THE INSPECTION OF NEW SAMPLES. BASED ON THIS INVENTION, MONITORING SAME PARAMETERS AS EXPECTED PERMEABILITY, DEFECTS OR DENSITY OF THE MATERIAL, THE PRODUCTION MANAGER HAS INFORMATION TO: DETERMINE THE VARIABILITY OF LOTS; IDENTIFY ANIMAL OBJECTS AND SEGREGATE THEM; PREPARING HOMOGENEOUS LOTS THROUGH THE SEPARATION OF INDIVIDUALS WITH DIFFERENT CHARACTERISTICS; MONITORING THE DEVELOPMENT OF NEW PROCESSES AND PRODUCTIVE PROCEDURES; MONITORING THE BEHAVIOR OF CORK OF DIFFERENT ORIGINS.

Description

1

DESCRIPTION

METHOD OF NON-DESTRUCTIVE AND NON-EVASIVE INSPECTION OF VEGETABLE MATERIALS BASED ON THE USE OF RADIATION AND LE C TROMAGNÉ TI CA " The present invention relates to a method of non-destructive and non-invasive inspection of plant materials based on the use of electromagnetic radiation comprising the steps of inspecting and classifying plant material, preferably cork stoppers. This classification is made in different categories of the same property such as the predicted permeability to gases, presence of defects or different densities. These parameters are obtained by signal processing collected from a non-invasive and non-destructive technique such as X-ray, from which a mathematical model is applied to the intended categorization. The mathematical model is calibrated by processing property data to be inspected on known samples. Said mathematical model is constructed on a previously selected set of samples, from the following variables: a matrix containing the value of the parameter to be calibrated - obtained by classical or standardized methods - and another matrix contains the signal obtained from the image of RX. The mathematical model is then used for the purpose of said categorization by the signal collected by the sensor in the inspection of new samples in the production process, and then the expected value of the desired parameter is calculated. The categorization can be done according to perspectives such as: (i) rejection of groups that do not meet the desired parameter through inspection and processing according to the adopted model (ii) acceptance of a group that meets the criteria established in the adopted model (iii) categorization in groups of estimated property (s)

BACKGROUND OF THE INVENTION The processing of cork and stoppers goes through continuous and judicious selection of visual quality, but this is essentially based on superficial appreciation by visual inspection or photographic means, and any internal structural defects remain to be clarified. This eliminates the total elimination of stoppers that do not fulfill their sealing function (liquids and gases). 3

On the other hand, although the cork industry has developed preventive and curative procedures in order to reduce the transfer of compounds with a sensory impact, namely chloroanisois residues, there has been no technical-scientific activity to improve or classify performance in terms of to the transfer of gases, for example to oxygen. The scientific literature often refers to the differences between cork stoppers and their comparison with alternative systems (Crochiere, 2007 / Kwiatkowski, 2007; Tran, 2007; Waters, 2007), but no attention has been paid to possible segregation or manipulation of the behavior to gas transfer, in particular in the case of cork stoppers.

The results of the most recent research on the filling of wine bottles are mainly focused on the following aspects: 1. the development of methods for measuring oxygen permeability 2. the comparative study of various wine bottle filling systems 3. The contribution of the obturation system to the management of oxygen in the basement, namely as a wine shelflife forecasting tool on the market

However, the applicants are not aware of any practical applications of the investigations carried out in order to improve the performance of cork stoppers.

On the other hand, in terms of the permeability measuring methods developed, non-destructive and non-invasive applications are not known, which, above all, are shown to result in a cork segregation tool, based on its permeability, prior to its own use .

With regard to the interpretation of gas diffusion and permeation phenomena, given the characteristics of cork, it is not feasible to apply the principles of conventional fluid dynamics where parameters such as density, pressure, temperature and particle velocity are well defined at a given point .

Given the porous structure, the oxygen transport regime is fundamentally conditioned by the mean free trajectory of the medium, and the use of statistical mechanics is recommended. It is within this scope that the novelty and relevance of the present invention is framed: 1. a database to classify the cork stopper in function of its structural and functional characteristics; 2. enable the identification of stoppers other than 1

Average distance traveled by a particle (photon, atom or molecule) between successive impacts (in European Nuclear Socíety) will allow an effective seal, namely the occurrence of structural defects; 3. allow the classification of stoppers with different oxygen permeabilities; 4. develop a classification methodology that allows them to be separated in order to place on the market lots with functional predictability.

Description of the invention

Cork cork framework and its function Cork with industrial application consists of a material of plant origin consisting of cells formed by the traumatic felongenus that covers the cork oak (Quercus suber). After their formation, these cells have walls / membranes and cytoplasm, but after cell death the cytoplasmic residues deposit on the inner side of the cell wall, being readsorbed or incorporated into the secondary wall (Amaral Fortes, 2004).

Given the high number of variables and parameters (factors inherent to climate, cell formation intra and interannual and still related to tree genetics) in game characteristic of a natural heterogeneous product, the predictability of cork stopper behavior and performance makes it is impossible, without using tools that holistically (and thus integrated) capture the maximum of heterogeneity / complexity, in a non-destructive way, for later classification. The activity of standardization of raw materials is therefore not an option for the sector. However, the increasing demand of users / consumers of cork stoppers, looking for a reliable and predictable product, leads to a very serious consideration of the development of techniques that allow the separation of individuals (corks) produced, according to the prediction of their behavior, in relation to parameters such as gas permeability, presence of defects (particularly those that are not visible to the surface) or density. Only in this way will it be possible to monitor the competition with alternative synthetic gaskets, whose technological nature is considerably easier to control.

Framing of the roll sector

Traditionally cork has been used in the form of cork as a seal for beverage bottles, and its use in wine preservation is its most important application. The natural cork stopper has unique physical properties, such as elasticity, hydrophobicity and impermeability to gases and liquids, that no other product, natural or artificial, has ever been able to match. These properties, in addition to an efficient seal, 7 ensure an adequate maturation or aging of the wine and hence its frequent use in high quality wines that are stored for long periods of time (Fischer, 1997; Insa, 2006).

This application has however been threatened due to the appearance of credible alternatives to this bottle filling system. Said alternative systems have been positioned in relation to the cork stopper as giving guarantees of sensorial and functional consistency, in particular as regards the transfer of compounds that can alter the organoleptic characteristics of that, as well as regularity and predictability of gaseous transfers between the of the bottle and the product contained therein.

In this way, it can be considered that the apparent inconsistency of the cork product is associated with considerable losses in the wine industry. On the other hand, the lack of an adequate technical response has allowed the same alternative shutter systems to conquer the market, to the detriment of the cork seal despite the innumerable technical, ecological and social advantages of this one. The desired functionality for wine bottle seals is through its sealing ability and its innocuousness and sensory neutrality.

In the case of the production of cork stoppers, it is based on a product of plant origin whose characteristics relate to a network of biological, genetic and soil-climatic factors. Industrial processing seeks on its side to obtain a product of constant quality, adopting productive practices that surpass the aforementioned variability and allow to provide a product with predictable characteristics as regards its capacity to preserve the quality of wine in bottles. It must therefore be borne in mind that, in the case of a natural product, the mastery of the technical aspects reaches its limit when it is necessary to accept the heterogeneity of the material itself, without prejudice to the general functionality sought, due to the lack of more powerful technological resources . The introduction of the technology described in this invention, based on information contained in cork and currently inaccessible, allows management of the production of cork stoppers based on objective parameters. The information gathered allows us to: • prepare predictions about the functional behavior of the bottle stopper in parameters such as permeability, presence of defects and density • reduce heterogeneity in a given batch by separating elements of different characteristics in the mentioned parameters • improve functional guarantees • to value the product commercially.

Relevance in the functionality of the cork sealant

One of the uncontrolled functions of cork sealants is their permeability to gases, especially to oxygen, and in particular from the perspective of their use in glass bottles. Oxygen plays a key role in the life of the product (wine), so its control on the packaging determines the longevity of the product on the market and is a critical factor for the wine sector. Indeed, " oxidative degradation " of wines quickly leads to a marked loss of sensory qualities of wine. This phenomenon translates, from the aromatic point of view, into a loss of characteristic aromas of new wines such as " floral " and " fruity " and, in parallel, by the appearance of own aromatic notes of older wines and / or atypical aromas associated with the deterioration of the product. From the chromatic point of view there is brownish color development called " non-enzymatic browning ". A number of studies point out, however, that aromatic deterioration precedes chromatic (WIDENRADT and SINGLETON 1974; and FERREIRA, et al, 1997). The shelf life of the bottled product will be conditioned by the combination of the resistance of the wine to oxidation (related to the composition of the wine in certain substances, in particular the phenolic compounds, and the pH at which it (1974), SINGLETON and KRAMLING (1976), and SINLGETON, (1987)) and the optimal amount of oxygen entering through the seal. However, the roll sector has not yet been able to keep up with this need, giving rise to the market growth of technological (synthetic) alternatives for the sealing of wine bottles.

This invention seeks to contribute to the classification of cork stoppers for example according to their permeability to oxygen, presence of defects and / or density. On the other hand it can also contribute to a significant improvement in the quality control methods applied in the roll sector.

Industrial Inspection

The applications of various electromagnetic regions, beginning in the visible, have developed since the nineteenth century for the most diverse technical-scientific purposes. These techniques are very useful in allowing the reading of objects with visually imperceptible details using non-destructive and non-invasive techniques of the material being analyzed. The principle of these applications is based on the fact that the electromagnetic energy, of an undulatory nature, undergoes interference, polarization, refraction, diffraction, reflection, among other effects, when crossing an object.

There are many applications of non-invasive and non-destructive methods based on the use of electromagnetic radiation, with a high incidence in medicine and for a variety of scientific and research purposes, and some industrial applications are also being developed. The possibility of obtaining non-destructive and non-destructive information is a current concern, which translates into the strong development of process analytical technology (PAT) for a wide range of areas requiring high levels of safety, security and control, for example in the pharmaceutical industry .

Although there are many analytical options available (eg NMR, ultrasound, ...) the limitations of its application at the functional level; i.e. in sample-detector array; and financially; i.e. investment and running costs; lead to industrial applications to rely more on the use of less expensive detection methods such as infra red and / or X-ray spectroscopy. The present patent is preferably based on the use of the latter radiation because it has the desired applicability, such as demonstrates its use in densitometry beyond the current field of technology and mature technology knowledge. The strategy described in the acquisition and processing of the signal may however be used with other signal acquisition techniques.

X-rays are electromagnetic emissions whose wavelength ranges from 5e-3 to 1nm, the energy of the photons being in the tens or hundreds of keV. This electromagnetic energy is generated by the transition of electrons in the atoms, or the deceleration of charged particles.

Like all electromagnetic energy of wave nature, the X-rays undergo interference, polarization, refraction, diffraction, reflection. Interference is used in medicine to obtain images whose information when contextualized in an anatomical plane allows the detection of anomalies. This principle of pattern detection in the internal structure of a tissue is not restricted to the clinical area being used generically in different objects of biological origin or not. Indeed the density of the material can be determined by the ratio of the radiation emitted to the absorbed radiation, an example being bone densitometry in medicine.

The most relevant aspects of this analysis are its non-destructive and non-invasive characteristics, enabling a high resolution structural study that can reach the molecular level, for example, molecular structure of RX. The complexity inherent in biological systems translates into a complete heterogeneity in the phenotype i.e. form and function of different tissues, as is the case of cork plant tissue. In this way its use for industrial purposes, which normally implies a high resolution classification at the level of specifications of physico-chemical properties, becomes highly difficult and susceptible to error. Thus obtaining this information in a non-invasive and non-destructive manner is an indispensable condition for strict classification of the object. The RX radiation for the aforementioned characteristics justify its selection as a methodology for the analysis and inspection of the biological material. In fact the images acquired by this process contain sufficient structural information for robust classification of the material.

The patterns visible in the 2D image with their gray gradients translate to differences in the density of the plant material. Darker areas correspond to lower resistance inversely brighter areas greater resistance to radiation.

This image is thus the projection of the 3D object into a 2D space corresponding to the " fingerprint of the respective radiation transfer resistance. The development of suitable equipment would allow on-line control " cork to cork stopper, similar to what is done today in screening the visual aspect, by image processing, thus allowing segmentation by flow subcategories of each natural cork. Mathematical signal processing. The present invention also describes the adoption of multivariate mathematical methods for signal processing obtained by a non-destructive and non-invasive spectrometric examination, with a view to the construction of algorithms that allow: 1. the detection of objects / stoppers for analysis; 2. recognition of a pattern recognition related to a given characteristic to be classified; 3. the classification of the same objects according to established criteria.

One of the challenges of the present invention lies at the level of mathematical processing allowing the decomposition of signal for the extraction of variables relevant for the determination of characteristics of cork stoppers impacting on their performance in the bottle, such as permeability, presence of defects and / or density, from a data set contained in a complex signal obtained by a non-invasive and non-destructive method such as, for example, X-ray analysis.

Industrial Inspection

The development of new measurement techniques that can be applied directly on samples, online, in particular in situations that require high standards of sequencing, control and qarantia of the products to be placed on the market, such as case of those linked to the pharmaceutical industry. The joint development of new mathematical techniques for analysis of frequency signals allows the use of spectroscopy for the simultaneous monitoring of a high range of materials from a single spectrum. The structure - micro and macro - determines the physical behavior of any material, namely mechanical properties, structural anomalies or transfer / resistance to mass transfer. The density of a material is a projection of its structure.

The techniques that allow the analysis of the structure are therefore a source of information indispensable to the decision and analysis of the structure of objects. The structural detector of choice is the X-ray, with applications in macro analysis and microstructure in addition to visual analyzes, including, in medicine, bone structure or other tissues, or in engineering the inspection of buildings and structures. In these applications, the density of the analyzed materials is the observed factor allowing simultaneously: • the identification of anomalies • the structural characterization • the interpretation of the microstructure

Industrial solutions applicable to the roll sector are not known for the purpose of classifying and separating stoppers according to their visual or structural characteristics, with the exception of methods based on the use of cameras to photograph the surface thereof, and the images processed for the purpose of their separation.

Said industrial systems, commonly referred to as the electronic choice of cork stoppers, are supplied by several manufacturers, with slight differences between the available products, mainly based on the number and positioning of the inspection chambers and software developed to facilitate human-machine communication .

In these equipments, the corks are separated from the same batch according to their external appearance and macroporosity, but the visual inspection performed by trained operators is not totally abandoned. In addition, these devices are only partially effective in eliminating structural defects that compromise cork performance. - 17 -

In the present invention, the preferred industrialization of the X-ray analysis technique is sought in order to predict, after signal processing, 1. the existence of structural defects that compromise the seal function of the stopper 2. the behavior to mass transfer and in particular to oxygen permeability. The fact that the radiographic image has a good correlation with the average free trajectory of the structure partly explains the establishment of the oxygen transport regime.

Mathematical signal processing The use of non-invasive and non-destructive signal acquisition methods, such as those based on the behavior of electromagnetic energy, leads to a large amount of input data. In general this database is too heavy for the elaboration of a practical application algorithm and contains a large amount of redundant information that does not contribute to the knowledge of the product or the process under analysis.

To compensate for this difficulty, we adopt mathematical manipulation to represent the information according to the features in a reduced / projected format. When extracting the information is performed according to carefully selected vectors, the data relevant to performing the desired analytical task are obtained, using the strictly necessary inputs and not the entire original information matrix, which is too complex.

That is, generically, the existing mathematical methods for the processing of spectra are based on the factorization of the spectra in a new base system (usually a sub-space) in order to obtain a representation with greater interpretability and to extract the parts of the spectrum with greater relevant systemic information.

This class of mathematical methods is described in great detail in the literature of the art under the topic of chemometrics ('chemiometrics'). The successful applications of these methods for analyzing chemical information and for obtaining quantitative models for the calibration of chemical species based on spectroscopy are well supported scientifically (some citations). Pattern extraction is therefore a methodology for selecting vectors that allow in practice to analyze a given problem with interesting accuracy. The use of techniques that allow the approximation of the selected vector patterns to the original database is essential. Such techniques allow to avoid redundant information and: 1. simplify the data to be processed; - 19 - 2. reduce the dimensions to be processed - by adopting latent variables; 3. systematize the information collected by the spectra.

This invention describes the application of signal decomposition techniques, supervised or not, respectively and as an example PLS (Partial least squares) and PCA (Principal components analysis), as techniques of multivariate analysis on the image recorded by a non-invasive methodology and non-destructive, for example X-ray. In this way it is possible to significantly reduce the number of attributes to be analyzed.

From the complex signal obtained from non-invasive and non-destructive spectroscopic analysis, the signal strictly necessary for classification is extracted: with the application of SVD (Singular Value Decomposition) we can highlight the relevant signal, however small, of the original noise. Selecting the appropriate preprocessing process is crucial to the end result. Another essential aspect of the development of the present patent is the need to identify patterns of behavior related to the variable to be quantified.

The data thus processed provides information that allows: automatic detection of objects - 20 - • calculation of regions of interest of the objects analyzed for later classification

BRIEF DESCRIPTION OF THE DRAWINGS The following description of a preferred embodiment is based on the accompanying drawings in which, without limitation, there is shown: - in Figure 1, a summary flowchart of the steps of the invention; - In figure 2, general diagram - algorithm development; In figure 3, a block diagram schematically representing the equipment for carrying out the method object of the invention; - In figure 4, steps in the detection of objects a - original image; b - noise reduction, subtraction of the fund and standardization; c - entropy-based binarization; d - contour detection; and - detection of circles by the Hough transform; and f - position and region of interest of the objects; In Figure 5, an RX image obtained after scanning the captured signal at the detector; In figure 6, the calibration curve of the detector; - In figure 7, image segmentation example for high brightness regions, a - original image, b - binarization of the image for regions of high brightness; In Figure 8, steps in obtaining the FFT of the growth vector, the Growth vector, b - brightness values, c - discrete Fourier transform (logarithmic scale); In Figures 8 and 9, correlation graphs between FFT and the value of the permeability of corks to oxygen are shown.

Listing of reference numbers of Figures 2 and 3 Figure 2

Image Processing (1)

Detector of Objects (2)

Processing of Objects (3)

Parameter prediction (4)

Classifier (5)

(6) Calculation of statistical parameters (7) FFT and Vector Growth (8)

Forecast Model (9)

Dataset Training (10)

X-ray image (11)

Object Position (12)

Calibration Parameters X-Ray / Detector (13) Object Classification (14)

Test object (15)

(16) Numerical Prediction Method (17) Laboratory Methods of Supervision (18)

Detector of defects (30)

Figure 3 - 22 -

Digital X-ray Detector (100)

Stoppers (200)

Carried on the carpet (300)

RX Emission Tube (400)

X-ray control unit (500)

I / O control unit (600)

Processing unit (700)

External information input / output to the system (800) Image transmission (900)

Programmable logic controller (1000)

Selector (1100)

DETAILED DESCRIPTION OF AN EMBODIMENT The invention described herein combines image analysis tools with a unique and innovative solution for sorting cork stoppers according to previously selected parameters such as gas permeability, presence of defects or density. The method for monitoring and supporting the management of cork stopper production is developed according to the steps shown in the flowchart shown in figures 1 and 2 and detailed below. The industrialization of the process necessarily involves: 1. the object detection algorithm; 2. the algorithm for detecting the growth vector on which signal processing is performed. The procedure for collecting and validating data requires the prior selection of samples representing the population for which the mathematical model is to be built for further automation of separation: for example stoppers with a certain size and with a certain surface treatment.

These samples are analyzed according to two complementary perspectives generating two datasets (figure D ·

Datasetl - Analysis by non-invasive and non-destructive technique, preferably X-ray, under pre-adjusted conditions.

Once an image is extracted with the desired sensitivity, the signal is mathematically processed. The projection of the structure into a plane (dimensional reduction) leads to an information loss. The Supervision method applied to the Forecast Model mitigates this information loss. The application of Fourier transform, hereinafter referred to as FFT, according to the vector 'years of growth' of the cork and maximum diameter of the object, allows later calibration with the relevant parameters. - 24 -

Figures 9 and 10 show the correlation graphs between FFT and the value of the permeability of corks to oxygen.

Dataset2 - Collection of relevant parameters, eg gas permeability, presence of defects or density.

As can be seen from figure 2, the interface components are as follows.

Inputs: - X-ray image (11) - Object position (12)

Manual inputs: - Test object (15). (17) - Parameter measured by supervisory method (18) laboratory normalized - Command stop (16)

Outputs: - Calibration Parameters X-Ray / Detector (13) - Object Classification (14)

Data Structures - 25 -

Calculation of Statistical Parameters (7)

Types of defects, mean values, median, fashion, standard deviation, etc. FFT and Vector Growth (8)

Discrete Fourier Transform (approximate result) of the original image and the growth vector. The result of the Fourier transform has dimension and cardinality equal to the original and complex values. Because the image values are real numbers, the result has a symmetry of values. In this way, the values of the matrix or vector can be taken as k.

Since cork stopper manufacturing processes do not ensure alignment of the phenomena under study - for example, patterns of two objects, though similar, may be out of phase - the phase information of the complex signal is considered as negligible.

For elimination of the phase information and considering each value of the matrix or vector in the form: a + bi the amplitude will be, A = a2 + b2

Forecast Model (9) - 26 -

Numerical prediction model (9) obtained by the supervisory laboratory method. This model will be represented by a function that, given a given growth vector, will assign a permeability value.

Dataset Training (10)

Historical record of growth vectors (obtained by the industrial method) and attribute class (obtained by the Laboratory Supervision Method).

Processes

Image Processing (1)

This block has the function of improving image quality from the Digital X-ray Detector (100), systematic error correction, noise reduction, contrast enhancement and normalization.

Detector of Objects (2)

Automatic algorithm for detecting objects in the radiographic image. The positions of the objects 12 may also be predetermined by external information. In this case, the self-detection mechanism is not executed. The object detection algorithm in the image / streaming of the radiographic detector follows the steps exemplified in Figures 4a-f and has an internal data structure which records the overall coordinates of each object, ensuring that each object is unique. The position of each object is obtained by maximum localities in the accumulator space generated by the Hough transform.

Processing of Objects (3)

Obtaining information relevant to the processing of each object such as: statistical parameters and FFT calculation.

Defect Detector (30)

Algorithm for detection of structural anomalies in the object.

Parameter prediction (4) (eg, permeability)

Numerical prediction algorithm, with supervision like PLS, of the parameter of the object (as the permeability to gases) according to Forecast Model.

Classifier (5) The final classification of each object is obtained by weighing the statistical parameters and by the expected quantification of the parameter to be determined (by permeability or by the presence of defects or predicted density).

(6) - 28 - Laboratory method of development and training of a numerical prediction model of the parameter to be determined (eg permeability, presence of defects or density), using Data Mining techniques.

Equipment

As shown schematically in the block diagram of Figure 3, the equipment necessary to carry out the method object of the invention essentially comprises an RX-emitting tube 400 and an X-ray digital detector 100. Between these two modules pass the corks 200 carried in the carpet 300, the image of which is collected after scanning the signal in the digital X-ray detector 100. The transmission of the image 900 is sent to the control unit 1/0 (600). The image is then processed and the stoppers are sorted into the processing unit (700) and sent to different reservoirs by the selector (1100). All input / output information outside the system (800) is represented in the diagram by the arrow. The drive of the conveyor belt (300) and the selector (1100) is carried out via the programmable automaton (1000).

Method Description

We can subdivide the system into two autonomous and competing processing blocks. The industrial process of inspection and classification, and the laboratory process. The classification process corresponds to steps (1) to (5) (Figure 2) and is intended to have a real-time performance. Real-time is understood as a process that obeys temporal constraints in agreement with the time constants of the associated process. The associated process will be the industrial production of cork stoppers. The process receives an image or video stream from the digital X-ray detector (100). From an image preprocessing (1), information for calibration (13) of the x-ray control unit (500) may be sent.

This is followed by the detection of new objects (2) and the calculation of the relevant statistical parameters (7) for study.

By external command one can force a given test object (15) to be sent for laboratory examination. After the laboratorial examination of the supervision method (18) to obtain the class attribute of a given object, the training dataset (10) and the forecast model (9) are updated by the updating process of the forecast model (6).

After prediction of parameter (4), in the permeability case, using the data structure and the forecast model (9), the final classification of the object (5) is obtained by adding the information of the structure (types of defects, density , statistical parameters, etc.). The final classification of the displayed object (14) is transmitted to the programmable controller (1000) for actuating the selector (1100).

Then, if the stop command 16 is not active, the command is sent to the automaton for driving the conveyor belt 300 and restarting the algorithm for the image processing process 1. Laboratory Methods of Supervision (18)

Standard laboratory or industrial methods are used to measure the parameter according to which the objects are categorized.

In the case of oxygen permeability predictions, several methods may be adopted, the methods being exemplary:

Electrochemicals

Colorimetric

Based on luminescence

With zirconium or micro GC sensors

Based on the promotion of reaction of oxygen with other molecules

Figure 7 shows an example of a radiographic image of a cork stopper. The most visible feature is the growth pattern. However, this is not the phenomenon of greater interest at this stage, but rather the anomalies, which do not present a regular pattern. The identification of defects in the object is obtained by segmenting the histogram of the image in a region of high brightness, region of regular patterns and in a region of low brightness. The segmentation points of the image are obtained by entropy criteria, thus achieving optimization of the histogram segmentation points and minimizing the number of false positives. The recognition of the different brightness areas is directly related to the potential quantification of a defect, which is the result of significant and intense differences in density. II - Determination of the density of the mass density from the optical density

This example demonstrates the possibility of determining the mass density, a parameter resulting from the macro and micro structural characteristics of the cork, from the image obtained by exposing the plant tissue to an RX detection system.

The critical optimization parameters to obtain the analytical signal, the average brightness of the image generated by the RX are: from X-ray analysis. • Power (A) • Voltage / Current (V) • Distance to object (m) • Exposure time (s)

This step was essential to confirm adherence of the technique to the desired analysis. The strong correlation between optical density and density of the samples supports the feasibility of the technique. 1. Equipment Calibration

Signal acquisition parameters: Power; Tension; Distance and Time:

The following RX ray equipment (of the Philips brand) was used with a digital detector of dimensions 200 x 300 mm, used to obtain the image. The standard curve is obtained using different thicknesses of aluminum foil material which within a thickness range (0.8 - 1.2 mm) has an optical density similar to cork plant material.

Simultaneously using the parameters of the first optimization cycle, RX images of corks from different classes are collected. The initial parameters were Power = 100 mA; Voltage = 40 kV; Distance = 1 m and Time = 0.1 sec

Parameters Image Scan:

RX images are obtained after scanning the RX signal captured on the detector. The import parameters in the scan were GA = 0,9; GS = 0.5; s = 70; L = 2.0. The image corresponding to each pattern is then imported using ImageJ Software. (see figure 5). The curve is then determined by abcissa the thickness of the aluminum foil and in ordinate the average brightness of the calculated image using its maximum area. The process is repeated iteratively until a linear curve with a higher correlation coefficient r > 0, 9.

In parallel, different RX images were collected from different samples of corks from different classes.

The average brightness of the object (stopper) is then determined by obtaining its aluminum thickness equivalent by interpolation.

Image Compatibility - Standardization - 34 -

By using the sheet whose RX transfer resistance is immediately lower than that of the cork study material, the brightness of the background of the image is adjusted by calculating the mode of the total image by adjusting to a value common to the whole image. Image normalization is performed by adjusting the brightness of the image under study to a pre-set value.

Elimination of Systemic Errors in the Image The acquisition of the radiographic image is accompanied by some deviations caused by factors that may be considered systematic. The variation of the ampoule temperature, the spatial non-uniformity of the sensitivity of the digital sensor, etc.

To eliminate the above, the image of the aluminum foil whose resistance to RX transfer is immediately inferior to the fashion of the study image was selected, subtracting the two.

Subtraction of Aluminum The final correction of the image is done by subtracting the respective RX image from corks (9 by 7 objects) less aluminum bottom 0.8 mm.

The stoppers are arranged horizontally in a matrix previously referenced for later identification; of 7 - 35 - column objects by 9 line objects. The calibration curve of the detector is determined with five standards prepared as follows: pP2mm 2mm aluminum foil; Pp4mm one 4mm aluminum foil; Pp6mm a 2mm aluminum foil overlapped with a 4mm sheet; Pp8mm two overlapping aluminum sheets; PplOmm two sheets of 4mm and one of 2mm. Thus, 32 sheets of 20.8 mm thick aluminum and 32 sheets of 4 mm were used. III - Permeability

The following examples relate to the development of methodology for predicting the permeability of corks to oxygen. • The oxygen permeability determination, as shown in the graphs of Figures 9 and 10, was used as the supervisory laboratory method (18) (Figures 2) for equipment meeting the ASTM standard.

The results obtained were referred to the end of the description of this example.

Vector Growth and Permeability Forecast

As previously described the radiographic image is collected, the predominant pattern in the radiographic image of the cork stopper is the natural growth pattern. This pattern also has the characteristic of varying essentially in one direction. We can therefore represent this pattern by a vector here called Vector Growth.

This vector can thus be a representation of the mechanical characteristics of the cork stopper, and where, besides the regular growth of cork, there is evidence of anomalies in this growth. From the direction of the lines formed by the growth pattern, the line perpendicular to these that passes through the center of the circle is calculated. The Vector Growth is thus obtained (Figure 8). The permeability prediction is obtained based on the FFT of the Vector Growth.

The results obtained and presented in the following examples and figures show a high correlation between FFT (Figures 9 and 10) and the permeability value itself measured with the supervision method (18), described in the algorithm (see diagram in the figure algorithm summary 2) .

In this example samples of natural cork stoppers were prepared which were subjected to: • X-ray analysis followed by the image processing described in this patent; • Determination of permeability by standard method. - 37 -

Bibliography ASTM F 1307-02; " Standard method for oxygen transfer rate through dry packages using a coulometric sensor "; EDT 2006

Amaral Fortes, M .; et al. ; 'Cork'; IST Press, 2a

Ed. 2004

Casey, J .; " Corking pressure: its causes and effects "; The Austr. Grapegrower & Winemaker, 402a, 1997, 34-38 Crochiere, G.K .; " Measuring Oxygen Ingress During Bottling / Storage ". Practical Winery & Vineyard, Jan / Feb 2007

Eleanor &Ray; " Heald Getting Closure. The continuing search for the best way to seal the bottle "; wine.appellationamerica.com/wine-review, January 18, 2008

Ferreira, A.C. et al .; " Identification of key Odorants related to the typical aroma of oxidation-spoiled white wines "; J. Agric. Food Chem. 2003, 51, 1377-1381 Ferreira, A.C. et al .; " Kinetics of oxidative degradation of white wines and how they are affected by selected technological parameters "; J. Agric. Food Chem. 2002, 50, 5919-5924

Fischer C., " Analysis of cork taint in wine and cork material at olfactory subthreshold level by solid phase microextraction, " Journal of Agricultural and Food

Chemistry, vol. 45, 1997.

Hart and Kleinig; '' The role of oxygen in the aging of bottled wine "; The Australian & New Zealand Grapegrower & winemaker, 2005 - 38 -

Insa, S., et al., &Quot; Assays on the simultaneous determination and elimination of chloroanols and chlorophenols from contaminated cork samples, " Journal of Chromatography Ar vol. 1122, pp. 215-221, 2006.

Kwiatkowski, M.J., et al; "The impact of closures, including screwcap with three different headspace volumes, on the compostion, color and sensory properties of a Cabernet Sauvignon wine for two years' storage "; Australian Journal of Grape and Wine Research, 13, 2007 Lopes, P. et al. . " Impact of storage position on oxygen through different closures into wine bottles, " Journal of Agricultural and Food Chemistry, 2006, 54 (18), 6741-6746

Lopes, P. et al. . " Nondestructive colometric method to determine the oxygen diffusion rate through closures used in winemaking, " Journal of Agricultural and Food Chemistry, 2005, 53, 6967-6973

Phillips, C .; " Recent Research: The Science of Closures, " From Wine Business Monthly, 09/15/2005; WBM, September 2005, www.winebusiness.com SINLGETON, (1987) ........... SINGLETON and KRAMLING (1976), .............

Taber, G.; To Cork or Not to Cork: Tradition, Romance, Science, and the Battle for the Wine Bottle - Scribner, 2007

Tran, T., et al. ; " Using Membrane Technology to Optimize Closure Performance ". The Australian & New Zealand Grapegrower & Winemaker, March 2007

Tudor, P .; " Is This The Closure For Your Wine An in-depth look at the pros and cons of wine development under screw caps. " Wine Business Monthly - July 2005

Waters, L., et al .; "Wine and Oxygen Research at the AWRI", Jan 2008, Sacramento WIDENRADT and SINGLETON, (1974)

WO 2007/026233 A2; Method and System of Multivariate Analysis on Slice-wise data of reference structure Normalized Images for Improved Quality in Positron Emission Tomography Studies

Lisbon, September 28, 2010

Claims (10)

Method of non-destructive and non-evasive inspection of plant materials based on the use of electromagnetic radiation for the selective and automatic separation of plant material based on a specific parameter described by an applicable mathematical model, characterized in that it comprises: signal acquisition; an automatic algorithm to detect the objects in the radiographic image; an algorithm for the detection of the growth vector on which signal processing is performed; - real-time signal processing; - software for the recognition of behavior patterns related to the variable to be quantified; - the categorization by means of the signal collected by sensors in the inspection of new samples. Method of non-destructive and non-fugitive inspection of plant materials according to claim 1, characterized in that the electromagnetic radiation is the X-ray. Method of non-destructive and non-evasive inspection of plant materials according to claim 1, characterized in that the mathematical model is built on a group of samples previously selected from the following variables: - a matrix containing the value of the parameter to calibrate; - an array containing the signal obtained from the R-X image and subsequently used for categorization by the signal collected by sensors to inspect new samples in the production process to calculate the expected value of the desired parameter. Method of non-destructive and non-fugitive inspection of plant materials according to claim 1, characterized in that the specific parameter to be quantified is gas permeability. Method of non-destructive and non-fugitive inspection of plant materials according to claim 4, characterized in that the gas whose permeability is to be measured is oxygen. Method of non-destructive and non-fugitive inspection of plant materials according to claim 1, characterized in that the specific parameter to be quantified is the presence of structures of different optical densities related to the presence of defects in the vegetable matter. Method of non-destructive and non-fugitive inspection of plant materials according to claim 1, characterized in that the plant material is cork. 3 Method of non-destructive and non-fugitive inspection of plant materials according to claim 1, characterized in that the cork material is stoppers. Method of non-destructive and non-invasive inspection of plant materials according to the preceding claims, characterized in that the categorization of the material involves: - rejection of groups that do not meet the desired parameter by inspection and processing according to the adopted model; acceptance of a group meeting the criteria laid down in the adopted model; and - categorization into groups of estimated property (s). Apparatus for carrying out the method according to the preceding claims, characterized in that it comprises: - an RX emitting tube (400) and a digital X-ray detector (100); between these two modules passes the corks 200 carried on the carpet 300, the image from which is collected after digitizing the signal in the digital X-ray detector 100; the image transmission (900) is sent to the control unit 1/0 (600); - subsequently the image is processed and the stoppers are sorted in the processing unit 700 and forwarded to different reservoirs by the selector - all information input / output to the system is represented in the diagram by the arrow 800; the drive of the conveyor belt 300 and the selector 1100 is performed through the programmable automaton 1000. Lisbon, September 28, 2010